Mapping and Quantifying the Human-Environment Interactions in Middle Egypt Using Machine Learning and Satellite Data Fusion Techniques

Abstract

Population growth in rural areas of Egypt is rapidly transforming the landscape. New cities are appearing in desert areas while existing cities and villages within the Nile floodplain are growing and pushing agricultural areas into the desert. To enable control and planning of the urban transformation, these rapid changes need to be mapped with high precision and frequency. Urban detection in rural areas in optical remote sensing is problematic when urban structures are built using the same materials as their surroundings. To overcome this limitation, we propose a multi-temporal classification approach based on satellite data fusion and artificial neural networks. We applied the proposed methodology to data of the Egyptian regions of El-Minya and part of Asyut governorates collected from 1998 until 2015. The produced multi-temporal land cover maps capture the evolution of the area and improve the urban detection of the European Space Agency (ESA) Climate Change Initiative Sentinel-2 Prototype Land Cover 20 m map of Africa and the Global Human Settlements Layer from the Joint Research Center (JRC). The extension of urban and agricultural areas increased over 65 km2 and 200 km2, respectively, during the entire period, with an accelerated increase analysed during the last period (2010–2015). Finally, we identified the trends in urban population density as well as the relationship between farmed and built-up land.